%define           ENABLE_SNMP_SUBAGENT 0

%define           ENABLE_MGMT 1

%define           uid 24

%define           gname haclient

%define           uname hacluster

Summary:          Heartbeat subsystem for High-Availability Linux

Name:             heartbeat

Version:          2.0.4

Release:          2%{?dist}

License:          GPL/LGPL

URL:              http://linux-ha.org/

Group:            System Environment/Daemons

Source0:          http://linux-ha.org/download/%{name}-%{version}.tar.gz

Patch0:           heartbeat-fedora-pam.patch

Patch1:           heartbeat-fedora-chkconfig.patch

Patch2:           heartbeat-fedora-ccmuser.patch

BuildRoot:        %{_tmppath}/%{name}-%{version}-%{release}-root-%(%{__id_u} -n) 

BuildRequires:    glib2-devel

BuildRequires:    iputils

BuildRequires:    openssl-devel

BuildRequires:    libnet-devel

BuildRequires:    libtool-ltdl-devel

BuildRequires:    net-snmp-devel

BuildRequires:    bzip2-devel 

%if %{ENABLE_MGMT}

BuildRequires:    gnutls-devel 

BuildRequires:    pam-devel

BuildRequires:    python-devel

BuildRequires:    swig

%endif

Requires:         stonith = %{version}-%{release}

Requires:         pils = %{version}-%{release}

Requires:         ldirectord = %{version}-%{release}

Requires:         pygtk2 >= 2.4

Requires(pre):    fedora-usermgmt

Requires(postun): fedora-usermgmt

Requires(post):   /sbin/chkconfig

Requires(preun):  /sbin/chkconfig

%if %{ENABLE_MGMT}

Requires:         gettext

%endif

%description

heartbeat is a basic high-availability subsystem for Linux-HA.

It will run scripts at initialization, and when machines go up or down.

This version will also perform IP address takeover using gratuitous ARPs.

It supports "n-node" clusters with significant capabilities for managing

resources and dependencies.

In addition it continues to support the older release 1 style of

2-node clustering.

It implements the following kinds of heartbeats:

        - Serial ports

        - UDP/IP multicast (ethernet, etc)

        - UDP/IP broadcast (ethernet, etc)

        - UDP/IP heartbeats

        - "ping" heartbeats (for routers, switches, etc.)

           (to be used for breaking ties in 2-node systems)

%package -n ldirectord

Summary:          Monitor daemon for maintaining high availability resources

Group:            System Environment/Daemons

Requires:         ipvsadm

Requires(post):   /sbin/chkconfig

Requires(preun):  /sbin/chkconfig

%description -n ldirectord

ldirectord is a stand-alone daemon to monitor services of real 

for virtual services provided by The Linux Virtual Server

(http://www.linuxvirtualserver.org/). It is simple to install 

and works with the heartbeat code (http://www.linux-ha.org/).

%package -n stonith

Requires:       pils = %{version}-%{release}

Summary:        Provides an interface to Shoot The Other Node In The Head 

Group:          System Environment/Daemons

%description -n stonith

The STONITH module (a.k.a. STOMITH) provides an extensible interface

for remotely powering down a node in the cluster.  The idea is quite simple:

When the software running on one machine wants to make sure another

machine in the cluster is not using a resource, pull the plug on the other 

machine. It's simple and reliable, albeit admittedly brutal.

%package -n pils

Summary:        Provides a general plugin and interface loading library

Group:          System Environment/Daemons

%description -n pils

PILS is an generalized and portable open source Plugin and Interface Loading 

System. PILS was developed as part of the Open Cluster Framework reference 

implementation, and is designed to be directly usable by a wide variety of 

other applications.

PILS manages both plugins (loadable objects), and the interfaces these plugins 

implement. PILS is designed to support any number of plugins implementing any 

number of interfaces.

%package devel 

Summary:        Heartbeat development package 

Group:          System Environment/Daemons

Requires:       heartbeat = %{version}-%{release}

%description devel

Heartbeat development package

%prep

%setup -q

%patch0 -p1

%patch1 -p1

%patch2 -p1

%build

# disable-fatal-warnings flag used to disable gcc4.x warnings of 'difference in signedness'

CFLAGS=${RPM_OPT_FLAGS} \

%configure \

  --disable-fatal-warnings \

  --disable-static \

%if %{ENABLE_MGMT}

  --enable-mgmt

%else

  --disable-mgmt

%endif

make

%install

rm -rf $RPM_BUILD_ROOT

mkdir -p $RPM_BUILD_ROOT

make DESTDIR=$RPM_BUILD_ROOT install

(

  cd $RPM_BUILD_ROOT/etc/ha.d/resource.d

  ln -s /usr/sbin/ldirectord ldirectord

)

# cleanup

[ -d $RPM_BUILD_ROOT/usr/man ] && rm -rf $RPM_BUILD_ROOT/usr/man

[ -d $RPM_BUILD_ROOT/usr/share/libtool ] && rm -rf $RPM_BUILD_ROOT/usr/share/libtool

find $RPM_BUILD_ROOT -type f -name *.la -exec rm -f {} ';'

sed -i -e '1i# -*-Shell-script-*-' $RPM_BUILD_ROOT/%{_libdir}/heartbeat/ocf-shellfuncs 

sed -i -e '1i# -*-Shell-script-*-' $RPM_BUILD_ROOT/%{_sysconfdir}/ha.d/shellfuncs

chmod -x $RPM_BUILD_ROOT/%{_libdir}/heartbeat/crmtest/helper.sh

chmod -x $RPM_BUILD_ROOT/%{_libdir}/heartbeat/ocf-shellfuncs

chmod -x $RPM_BUILD_ROOT/%{_sysconfdir}/ha.d/shellfuncs

chmod -x $RPM_BUILD_ROOT/%{_libdir}/heartbeat/pymgmt.py

%clean

if [ -n "${RPM_BUILD_ROOT}"  -a "${RPM_BUILD_ROOT}" != "/" ]; then

  rm -rf $RPM_BUILD_ROOT

fi

rm -rf $RPM_BUILD_DIR/heartbeat-%{version}

%pre

/usr/sbin/fedora-groupadd %{uid} -r %{gname} &>/dev/null || :

/usr/sbin/fedora-useradd %{uid} -r -s /sbin/nologin -d /var/lib/heartbeat/cores/hacluster -M \

        -c 'heartbeat user' -g %{gname} %{uname} &>/dev/null || :

%post

/sbin/ldconfig

/sbin/chkconfig --add heartbeat

%preun

/sbin/chkconfig --del heartbeat

%postun 

/sbin/ldconfig

test "$1" != 0 || /usr/sbin/fedora-userdel  %{uname} &>/dev/null || :

test "$1" != 0 || /usr/sbin/fedora-groupdel %{gname} &>/dev/null || :

%post -n ldirectord

/sbin/chkconfig --add ldirectord

%preun -n ldirectord

/sbin/chkconfig --del ldirectord

%post -n stonith -p /sbin/ldconfig

%postun -n stonith -p /sbin/ldconfig

%files

%doc %{_datadir}/doc/%{name}-%{version}

%defattr(-,root,root)

%dir %{_sysconfdir}/ha.d

%{_sysconfdir}/ha.d/harc

%config(noreplace) %{_sysconfdir}/ha.d/shellfuncs

%{_sysconfdir}/ha.d/rc.d

%config(noreplace) %{_sysconfdir}/ha.d/README.config

%{_libdir}/heartbeat

%{_libdir}/libapphb.so.*

%{_libdir}/libccmclient.so.*

%{_libdir}/libcib.so.*

%{_libdir}/libclm.so.*

%{_libdir}/libcrmcommon.so.*

%{_libdir}/libhbclient.so.*

%{_libdir}/liblrm.so.*

%{_libdir}/libpengine.so.*

%{_libdir}/libplumb.so.*

%{_libdir}/libplumbgpl.so.*

%{_libdir}/librecoverymgr.so.*

%{_libdir}/libstonithd.so.*

%{_prefix}/lib/ocf

%{_sysconfdir}/ha.d/resource.d/

%exclude %{_sysconfdir}/ha.d/resource.d/ldirectord

%{_sysconfdir}/init.d/heartbeat

%config(noreplace) %{_sysconfdir}/logrotate.d/heartbeat

%dir %{_var}/lib/heartbeat

%dir %{_var}/lib/heartbeat/cores

%dir %attr (0755, root, root) %{_var}/lib/heartbeat/cores/root

%dir %attr (0755, nobody, nobody) %{_var}/lib/heartbeat/cores/nobody

%dir %attr (0755, hacluster, haclient) %{_var}/lib/heartbeat/cores/hacluster

%dir %{_var}/run/heartbeat

%attr (0755, hacluster, haclient) %{_bindir}/cl_status

%{_bindir}/cl_respawn

%{_sbindir}/crmadmin

%{_sbindir}/cibadmin

%{_sbindir}/ccm_tool

%{_sbindir}/crm_diff

%{_sbindir}/crm_uuid

%{_sbindir}/crm_mon

%{_sbindir}/iso8601

%{_sbindir}/crm_master

%{_sbindir}/crm_standby

%{_sbindir}/crm_attribute

%{_sbindir}/crm_resource

%{_sbindir}/crm_verify

%dir %attr (755, hacluster, haclient) %{_var}/run/heartbeat/ccm

%dir %attr (755, hacluster, haclient) %{_var}/run/heartbeat/crm

%dir %attr (755, hacluster, haclient) %{_var}/lib/heartbeat/crm

%{_mandir}/man1/cl_status.1*

%{_mandir}/man1/ha_logger.1*

%{_mandir}/man1/hb_standby.1*

%{_mandir}/man1/hb_takeover.1*

%{_mandir}/man1/hb_addnode.1*

%{_mandir}/man1/hb_delnode.1*

%{_mandir}/man8/heartbeat.8*

%{_mandir}/man8/apphbd.8*

%{_mandir}/man8/ha_logd.8*

%if %{ENABLE_SNMP_SUBAGENT}

/LINUX-HA-MIB.mib

%endif

%if %{ENABLE_MGMT}

%{_libdir}/libhbmgmt.so.*

%{_libdir}/libhbmgmtclient.so.*

%{_libdir}/libhbmgmtcommon.so.*

%{_libdir}/libhbmgmttls.so.*

%config(noreplace) %{_sysconfdir}/pam.d/hbmgmtd

%{_datadir}/locale/zh_CN/LC_MESSAGES/haclient.zh_CN.mo

%endif

%files -n ldirectord

%doc doc/COPYING 

%doc doc/README

%doc ldirectord/ldirectord.cf

%defattr(-,root,root)

%{_sysconfdir}/ha.d/conf

%{_sbindir}/ldirectord

%{_sbindir}/supervise-ldirectord-config

%config(noreplace) %{_sysconfdir}/logrotate.d/ldirectord

%{_sysconfdir}/init.d/ldirectord

%{_sysconfdir}/ha.d/resource.d/ldirectord

%{_mandir}/man8/ldirectord.8*

%{_mandir}/man8/supervise-ldirectord-config.8*

%files -n stonith

%doc doc/COPYING 

%doc doc/README

%defattr(-,root,root)

%{_libdir}/*.so.*

%{_libdir}/stonith/

%{_sbindir}/stonith

%{_sbindir}/meatclient

%{_mandir}/man8/stonith.8*

%{_mandir}/man8/meatclient.8*

%files -n pils

%doc doc/COPYING 

%doc doc/README

%defattr(-,root,root)

%{_libdir}/pils/

%files devel

%doc %{_datadir}/doc/%{name}-%{version}

%defattr(-,root,root)

%{_includedir}/heartbeat/

%{_includedir}/clplumbing/

%{_includedir}/saf/

%{_includedir}/ocf/

%{_includedir}/stonith/

%{_includedir}/pils/

%{_libdir}/*.so

%changelog

* Wed Mar 29 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.4-2

- Version 2.0.4

* Wed Mar  1 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-9

- changed user creation

- added patch2 heartbeat-2.0.3-fedora-ccmuser.patch  

* Wed Mar  1 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-8

- specifically excluded ldirectord symlink from heartbeat package

- removed user and group deletion in postun

- renamed subpackages ldirectord, pils and stonith to lose prefix heartbeat

  by using -n

* Tue Feb 28 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-7

- fixed more rpmlint errors and warnings

* Sat Feb 25 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-6

- fixed number of rpmlint warnings and errors (still ignores some..)

- generate 'predictable' uid and gid with fedora-usermgmt to use with 

  configure flag -with-ccmuser-id and groupadd, useradd

- added Buildreq's: libtool-ltdl-devel, fedora-usermgmt-setup

  net-snmp-devel, bzip2-devel 

- removed *.so duplication in heartbeat and heartbeat-devel

- changed file sections

* Fri Feb 24 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-5

- useradd with fedora-usermgmt

- added *.so file to -devel sub-package

* Sat Feb 18 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-4

- removed all perl requires; should be picked up by rpmbuild automagically

- changed system user creation hacluster part to use baseid:

  (http://fedoraproject.org/wiki/Packaging/UserCreation)

* Thu Feb 16 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-3

- removed Requires: python and gnutls

- changed _libdir/ocf -> _prefix/lib/ocf

- reversed subpackages depend on basepackage

- removed Req swig (kept BuildReq)

- added Req pygtk2

* Wed Feb 15 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-2

- fixes for various rpmlint errors and warnings 

- fixed setup -q

- make subpackages depend on basepackage, not reverse

- clean buildroot at beginning of install

- replaced a number of hardcoded paths with RPM macros

- Changed Group from Networking/Daemons to System Environment/Daemons

- enable mgmt option

* Sun Feb 12 2006  Joost Soeterbroek <fedora@soeterbroek.com> - 2.0.3-1

- rebuilt for Fedora Extras

* Fri Feb 10 2006  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 2.0.3 - Bug fixes and significant new features.

  + Management Daemon/Library and GUI client

    + provide a management library for manamgement daemon and CIM provider

    + provide a management daemon and a basic GUI management tool

  + CIM enablement

    + CIM (Common Information Model) enablement - works with

      sblim-sfcb, OpenWBEM, and Pegasus CIMOMs

    - not yet compiled into our binary RPMs because of dependencies

  + CRM (Cluster Resource Manager) General

    + All shutdowns go via the PE/TE - preserves inter-resource ordering

    + Support for future changes to the CIB (depreciation of cib_fragment)

    + Overhaul of IPC and HA channel callback logic

    + Many improvments to the quality and quantity (reduced) of logging

  + CRMd

    + Timerless elections - when everyone has voted we're done

    + Use the replace notification from the CIB to re-update our copy with 

      our view of our peers.

    + Reliably detect if the LRM connection is still active.

    + Elections

      + newer versions defer to older ones in DC elections 

        (opposite of current behavior)

      + this means that only once the complete cluster has been upgraded will

        we start acting like the new version and accept new config options

      + it also means newer PE's and TE's (the most complex pieces) don't need

        "act like the old version" options and can rely on all slaves being at

        least as up-to-date as they are

      + people can run mixed clusters as long as they want

        (until they want the new PE features)

      + new DCs only update the version number in the CIB if they have a 

        higher value

      + nodes that start and have a lower version than that stored in the CIB

        shut themselves down (the CRM part anyway)

      + this prevents an admin from introducing old nodes back into an upgraded

        cluster. It probably doesn't fully understand the config and may not

        support the actions the PE/TE requires.

  + CIB (Common Information Base daemon)

    + Make sure "query only" connections cant modify the CIB

    + Periodically dump some stats about what the CIB has been doing.

    + Verify there are no memory leaks

    + Performance enhancements

    + Prevent a single CIB client from blocking everyone else

    + Clients Can be notified of full CIB replacements

    + record_config_changes option in ha.cf for those worried about 

      the amount of logging.  Defaults to "on".

    + suppress_cib_writes CIB option replaced with in enable_config_writes ha.cf 

      (enable_config_writes to be removed in 2.0.4)

    + Never write the status section to disk

    + Check permissions for the on-disk CIB at startup

    + Dont trash unreadable on-disk CIBs

    + Fix for updates made against the whole CIB (not just one section) 

  + PEngine (Policy Engine)

    + Many improvements to the handling of resource groups

    + Support "anonymous" clones

    + Fix stonith ordering

    + Order DC shutdowns after everyone else's

    + Support short resource names (for group and clone resources)

    + The ordering and colocation of grouped resources is now optional

    + Support probing new nodes for active resources.

    + All "probe" actions are controlled by the PE.

      + No resource may be started until the probing is complete.

      + Do not probe for resources we know to be active on unprobed nodes

    + When looking for monitor ops, only mark it optional if it was already

      active on the node we're interested in.

    + Detect changes to class/type/provider/parameters and force a restart

      of the resource

    + New record_pengine_inputs option in ha.cf for those worried about 

      the amount of logging.  Defaults to "on".

    + Differentiate between config and processing errors

      + reduces the frequency that we need to log the complete CIB

    + Make notify for master/slave work

    + New CIB option: stop_orphan_actions (boolean)

      If a resource is no longer defined, we can optionally stop it

    + New CIB option: stop_orphan_actions (boolean)

      If a monitor op for a given interval is no longer defined, we can

      optionally stop it

    + Add support for time and phase-of-the-moon based constraints

    + Improved failure handling: avoiding false positives

    + Always create orphaned resources - so they show up in crm_mon

    + Do not require sequential clone numbers starting at 0

  + TEngine (transition engine)

    + Detect old stonith ops

  + CLIs (Command Line interfaces)

    + Create a --one-shot option for crm_mon

    + Switch a number of CLI tools to use the new syncronous connections

    + Log errors to stderr where they will be seen and therefore useful

    + Support migration and un-migration of resources and resource groups

    + Create crm_verify for checking configuration validity

    + Simplify the passing of XML to cibadmin

  + Known open bugs worth mentioning:

    + 1075, 1080, 1081, 1084, 1085, 1064, 1069, 756, 984

    + 1050, 1082, 1037, 1079

* Thu Sep 22 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 2.0.2 - small bug fix only release

  + Fixed a bug in ping directive - it works again

  + Added a check to BasicSanityCheck to check ping and ping_group directives

  + fixed cl_status nodestatus to return 0 if a node has status "ping"

  + fixed a memory leak in the CRM's LRM interface code

  + fixed code which deterimines which version of the CRM becomes

    the DC when basic CIB schema versions differ.  It now prefers

    the older version to be DC instead of the newer version.

* Wed Sep 14 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 2.0.1 - 

  + Communication Layer

    + netstring encoding format is changed to be more efficient

    + add compression capability for big messages

  + Add man pages for hb_standby/hb_takeover	

  + The assert triggered by 2.0.0 has been fixed

  + CIB can now contain XML comments and/or be in DOS format	

  + Includes implementation of the ISO8601 date format

  + New CLI tools for changing cluster preferences, node attributes 

    and node standby

  + Improved recovery and placement of group resources

  + Detection of failed nodes by the Policy Engine is fixed

  + New Policy Engine features 

    http://www.linux-ha.org/ClusterResourceManager/DTD1.0/Annotated :

      sections 1.5.[8,9,10,12]

    + Constraints and instance attributes can now be active conditionally

    + Rules can now contain other rules

    + Date/Time based expressions are supported

    + Cloned resources can now optionally be notified before and after

      any of its peers are stopped or started.

    + The cluster can re-evaluate the configuration automatically after

      a defined interval of idleness

  + Removed a flow control message which was very annoying when operating

    in a mixed 1.x/2.x environment

  -- Known Bugs :-( --

    - Bug 859 - FSA took too long to complete action - fully recovered from

    - Bug 882 - IPC channel not connected during shutdown - harmless

    - Bug 879 - Failed actions cause extra election - harmless

 Each of these occurs about once or twice in 5000 test iterations

       - This is probably > 10K failovers

    - rsc_location constraints cannot have rules that contain other rules

      (fixed in CVS after release) 

* Fri Jul 29 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 2.0.0 - First stable release of the next generation of the Linux-HA project

  + Basic Characteristics described here:

	http://linux-ha.org/FactSheetv2

  + Core infrastructure improvments:

    + Messaging (message acknowledging and flow control)

    + Logging   (logging daemon)

  + Release 1.x style (2-node) clusters fully supported

  + Multi-node support (so far up to 16-node clusters tested)

	See http://linux-ha.org/GettingStartedV2 for more information

  + New components:

    + Cluster Information Base    (replicated resource configuration)

    + Cluster Resource Manager    (supporting 1->N nodes)

    + Modular Policy Engine       (controlling resource placement)

    + Local Resource Manager      (policy free, not cluster aware)

    + Stonith Daemon              (stand-alone fencing  subsytem)

  + Support for OCF and LSB resource agents

  + Support for composite resource types (groups, clones)

  + Support for a rich set of resource location and ordering constraints

  + Conversion tool for existing haresources

  + Resources monitored by request

  + Resource "maintenance" mode

  + Several failback, failure and "No Quorum" behaviours to choose from

        (global defaults and per action or resource)

  + Sample cluster state and configuration monitoring tools

  Known issues in 2.0.0:

    - Under some rare circumstances the cluster manager will time out

      while stabilizing a new cluster state.  This appears to be

	otherwise harmless - the cluster is actually fine.

	http://www.osdl.org/developer_bugzilla/show_bug.cgi?id=770

    - Under some rare circumstances, a dev assert will be triggered

	in unpack.c.  This results in the pengine getting restarted.

	This is annoying, but not a disaster.

	http://www.osdl.org/developer_bugzilla/show_bug.cgi?id=797

* Tue May 23 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.99.5 - Near-final beta of 2.0.0 release

  + many bug fixes - code looks very stable at this point

    -- well tested at this point on 4 and 8 node clusters.

* Thu Apr 07 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.99.4 - Near-final beta of 2.0.0 release

  + many bug fixes since 1.99.1

  + new external STONITH model - fully supports scripting interface

  + tested through 12 node clusters successfully

  + No serious defects found in testing

  + Easier-to-understand locational constraints model

  + Many bug fixes of many kinds

  + Important bug fixes to OCF IPaddr resource agent

  + Resources are monitored only on request

  + See http://wiki.linux-ha.org/ClusterResourceManager/Setup

    for basic ideas about getting started.

  + Release 1 style (2-node) clusters still fully supported

  + Release 2 style clusters support 1-N node clusters

	(where N is probably something like 8-32)

* Tue Mar 20 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.99.3 - Near-final beta "technology preview" of 2.0.0 release

  + many bug fixes since 1.99.1

  + tested through 12 node clusters with reasonable success

  + new STONITH API

* Sun Feb 20 2005  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.99.2 - Near-final beta "technology preview" of 2.0.0 release

  + Many many many changes.  Far too many to describe here.

  + See http://wiki.linux-ha.org/ClusterResourceManager/Setup

    for certain basic ideas about getting started.

* Mon Oct 11 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.99.1 - *early* beta series - preparing for 2.0.0

  + Andrew provided a number of fixes to the CRM and 2.0 features

  + Fixed a problem with retrying failed STONITH operations

* Mon Oct 11 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.99.0 - *early* beta series - preparing for 2.0.0

  + All STABLE changes noted below have been ported to this branch

  + Included in this release is a beta of the next generation of Heartbeat

	resource manager developed by Andrew Beekhof.  

	http://linuxha.trick.ca/NewHeartbeatDesign is a good place to learn

	more about this effort. Please examine crm/README, crm/test/README

	and crm/crm-1.0.dtd for example usage and configuration.

  + Also included is the L(ocal) R(esource) M(anager) developed by IBM China

	which is an integral part of the NewHeartbeatDesign.

  + Known caveats:

    - STONITH as a whole has seen a code cleanup and should be tested

      carefully.

    - The external STONITH plug-in has undergone major surgery and

      probably doesn't work yet.

    - the new CRM is not perfectly stable with 3 nodes yet.

  + PLEASE see http://osdl.org/developer_bugzilla/enter_bug.cgi?product=Linux-HA

    and use it to report quirks and issues you find!

* Sat Sep 18 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.2.3 (stable)

  + fixed a serious error which causes heartbeat to misbehave after about

        10 months of continuous operation

  + Made our ARP packets more RFC compliant

  + Extended apcmastersnmp code to deal with new devices

  + fixed a bug concerning simultaneous stops of both machines causing one

        of them to not shut down.

  + added an option to suppress reporting on packet corruption

  + fixed it so that we don't create the FIFO by the RPM

  + made cl_status setgid so anyone can run it, and fixed exit codes

  + eliminated a serious memory leak associated with client code

  + packaged doc files which had been missed before

  + fixed many many small bugs and memory leaks detected by BEAM

  + added several new test cases

  + fixed longstanding bug in plugin unloading

  + fixed a shutdown hang problem

  + several fixes for Solaris, FreeBSD

  + Solaris packaging now included in base

  + fixed a bug related to the apache resource agent not handling

        quoted parameters

  + added use_apphbd parameter to have heartbeat register

        with apphbd instead of watchdog device when desired

  + changed apphbd to default its config file to /etc

  + added snmp subagent code

  + added hbaping communications plugin

  + added external STONITH plugin

  + ldirectord: fixed a bug where real servers that were are

        present in multiple virtual services will only be added

        to one virtual service.

* Mon May 11 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.2.2 (stable)

  + Fixed several format string errors in communication plugins

  + Fixed a bug which kept us from diagnosing errors in non-aliased interfaces

  + Fixed a bug in ipaddr which caused an infinite loop when auto_failback on

  + Updated Debian things...

  + Added IPv6addr resource agent

  + Added ibmhmc STONITH plugin

  + Added cl_status command

  + Fixed a bug regarding restarts when auto_failback is on...

  + Fixed a couple of bugs in sha1 authentication method for very long keys

  + Fixed a bug in the portblock resource agent so that it no longer blocks

          ports on the loopback interface

  + Increased the time allowed for split brain test before it declares failure

+ Version 1.2.1 (stable)

  + Netstrings can now be used for our on-the-wire data format

  + Perl/SWIG bindings added for some heartbeat libraries

  + Significant improvements to SAF data checkpointing API

  + Implemented unbuffered ipcsocket code for SAF APIs

  + Many Solaris fixes -- except for ipfail, Solaris works

  + Significant library restructuring

  + Watchdog device NOWAYOUT is now overridded if defaulted

  + Watchdog device now kills machine instantly after deadtime

        instead of after one minute

  + Hostnames should now be treated case-independently...

  + Added new client status APIs - client_status() and cstatus_callback()

  + Fixed bug with auto_failback and quick full restarts

  + We now automatically reboot when resources fail to stop correctly...

  + We now check the status of the configured STONITH device hourly...

  + STONITH operations repeat after a 5 second delay, not immediately...

  + Added hb_takeover command - complement to hb_standby

  + Added documentation on how to use evlog/TCP to enable testing to

        take place without losing messages due to UDP message forwarding

  + Several new tests from Mi, Jun - split brain, bandwidth, failure

        detection time.

  + Fix to LVM resource from Harald Milz <hm@muc.de>

  + Fixed FreeBSD authentication problems breaking ipfail

  + Fixed .so loading on Debian

  + Fixed false complaints about resource scripts (from Jens Schmalzing)

  + Fixed false stop failure from LinuxSCSI  (from Jens Schmalzing <j.s@lmu.de>)

* Thu Apr 15 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.3.0 - beta series

  + Netstrings can now be used for our on-the-wire data format

  + Perl/SWIG bindings added for some heartbeat libraries

  + Significant improvements to SAF data checkpointing API

  + Implemented unbuffered ipcsocket code for SAF APIs

  + Many Solaris fixes -- except for ipfail, Solaris works

  + Significant library restructuring

  + Watchdog device NOWAYOUT is now overridded if defaulted

  + Watchdog device now kills machine instantly after deadtime

 	instead of after one minute

  + Hostnames should now be treated case-independently...

  + Added new client status APIs - client_status() and cstatus_callback()

  + Fixed bug with auto_failback and quick full restarts

  + We now automatically reboot when resources fail to stop correctly...

  + We now check the status of the configured STONITH device hourly...

  + STONITH operations repeat after a 5 second delay, not immediately...

  + Added hb_takeover command - complement to hb_standby

  + Added documentation on how to use evlog/TCP to enable testing to

	take place without losing messages due to UDP message forwarding

  + Several new tests from Mi, Jun - split brain, bandwidth, failure

	detection time.

  + Fix to LVM resource from Harald Milz <hm@muc.de>

* Tue Feb 16 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.2.0

  + Replaced the nice_failback option with the auto_failback option.

	THIS OBSOLETES THE NICE_FAILBACK OPTION. READ THE DOCS FOR HOW

	TO UPGRADE SMOOTHLY.

  + Added a new feature to hb_standby which allows you to give up

	  any specific category of resources:  local, foreign, or all.

	The old behavior is "all" which is the default.

	This allows you to put a auto_failback no cluster into

	  an active/active configuration on demand.

  + ipfail now works properly with auto_failback on (active/active)

  + ipfail now has "hysteresis" so that it doesn't respond immediately

	to a network failure, but waits a little while so that the

	damage can be properly assessed and extraneous takeovers avoided

  + Added new ping node timeout directive "deadping"

  + Made sure heartbeat preallocated stack and heap, and printed a

	message if we allocate heap once we're started up...

  + IPMILan STONITH plugin added to CVS

  + Added IPaddr2 resource script

  + Made the APC smart UPS ups code compatible with more UPSes

  + Added a (preliminary?) ordered messaging facility from Yi Zhu

  + Changed IPaddr's method of doing ARPs in background so that

	certain timing windows were closed.

  + Added OCF (wrapper) resource script

  + Allow respawn programs to take arguments

  + Added pinggroups (where any node being up is OK)

  + SIGNIFICANT amount of internal rearchitecture.

  + Many bug fixes.

  + Several documentation updates.

* Tue Feb 10 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.1.5

  + ipfail now has "hysteresis" so that it doesn't respond immediately

	to a network failure, but waits a little while so that the

	damage can be properly assessed and extraneous takeovers avoided

  + Several fixes to cl_poll()

  + More fixes to the IPC code - especially handling data reception

	after EOF

  + removed some unclean code from GSource for treating EOF conditions

  + Several bugs concerning hanging when shutting down early during startup

  + A few BasicSanityCheck bug fixes

  + CTS now allows a single machine to be able to monitor several clusters

  + Most former CTS options are now either unneeded or on the command line

  + Increased number of ARPs and how long they're being sent out

  + Fixed uncommon (authorization) memory leak

  + Some Solaris portability fixes.

  + Made init script handle standby correctly for new config files

  + Improved the fast failure detection test

  + Added some backwards compatibility for nice_failback and some default

	authentication directives

  + Corrected the 1.1.4 change log

* Fri Jan 22 2004  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.1.4

  + ipfail now works properly with auto_failback on (active/active)

  + Changed the API to use sockets (IPC library) instead of FIFOs.

  + Added new apiauth directives to provide authorization information

	formerly provided by the FIFO permissions.

  + Added Intel's implementation of the SAF data checkpointing API and daemon

  + Added a cleanup suggested by Emily Ratliff.

  + IPMILan STONITH plugin added to CVS

  + Added IPaddr2 resource script

  + Various cleanups due to horms.

  + Fixed authentication to work on 64-bit platforms(!)

  + Fixed the cl_poll() code to handle corner cases better

  + Made heartbeat close watchdog device before re-execing itself

  + New CTS improvements from Mi, Jun <jun.mi@intel.com>

  + Various minor bug fixes.

      . Several shutdown bugs addressed

      . fixed sendarp to make a pid file, so we can shut it down

          when we shut everything else down in case it's still running.

      . Lots of minor bug fixes to IPC code

      . Lots of minor bug fixes to ipctest program

      . made BasicSanityCheck more tolerant of delays

      . Fixed IPC code to authenticate based on ints, not int*s.

      . Check properly for strnlen instead of strlen...

      . Several signed/unsigned fixes

      . A few uninitialized vars now are inited

      . Switched to compiling lex/yacc sources the automake way

      . Lots of minor CTS fixes...

  + ldirectord bug fixes:

    . When new real servers are added on initialisation or when

        the configuration file is reread they are marked with status

        of -1 (uninitialised) so they will be checked and inserted

        into the virtual service as required

    . All checks use the checkport if set, otherwise the port set for

        the individual real server. This was the case for http and

        connect checks, but others had variations on this theme.

    . When the configuration file is reread because it changed

        on disk and autoreload is set, check the real servers

        immediately rather than waiting for checkinterval to expire

    . Already running message sent to stderr instead of stdout

    . Support alternate server in real-server specific URL

    . Treat the same real server with different weights as a different

        real server. Fixes bug reported by Philip Hayward whereby the same

        real-server would always have the same weight, regardless of

        the ldirectord.cf

* Fri Sep 26 2003  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.1.3

  + Bugfix for heartbeat starting resources twice concurrently if

    auto_failback was set to "legacy".

  + Bugfix for messages getting lost if messages were sent in quick

    succession. (Kurosawa Takahiro)

  + Bugfix for Filesystem resource checking for presence of filesystem

    support before loading the module.

  + BasicSanityCheck extended to cover more basic tests.

  + Bugfix for findif not working correctly for CIDR netmasks.

  + Minor bugfix for ldirectord recognizing new schedulers correctly and

    timeout settings are now being honoured.

  + Enhanced the message giving a better explanation of how to set up node

    names properly when current node not found in the ha.cf file

  + Send a message to the cluster whenever we have a node which doesn't

    need STONITHing - even though it's gone down.  This fix needed

    by CCM, which is in turn needed by EVMS.

  + Enhanced the messages for missing ha.cf and missing haresources files

    explaining that sample config files are found in the documentation. 

  + Fix for memory leak from Forrest Zhao<forrest.zhao@intel.com>

  + Added a (preliminary?) ordered messaging facility from Yi Zhu

  + FAQ updates

  + Added Xinetd resource script

  + Added OCF (wrapper) resource script

  + Allow respawn programs to take arguments

  + Added pinggroups (where any node being up is OK)

  + fixed ldirectord negotiatetimeout for HTTP

  + fixed a bug which caused -d flag to be ignored

  + failing resource scripts are now ERRORs not WARNings

  + now shuts down correctly when auto_failback == legacy

* Mon Jul 13 2003  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.1.2

  + Replaced the nice_failback option with the auto_failback option.

	THIS OBSOLETES THE NICE_FAILBACK OPTION. READ THE DOCS FOR HOW

	TO UPGRADE SMOOTHLY.

  + Changed IPaddr to not do ARPs in background, and shortened time 

	between ARPs.  Also made these things tunable...

  + changed our comm ttys to not become our controlling TTYs

  + Enhanced the ServeRAID script to fix a critical bug by using a new feature

  + Added a new DirectoryMap to CVS - tells where everything is...

  + significantly enhanced the BasicSanityCheck script, and the tests

	it calls.

  + added a new option to use a replacement poll function for improved

	real-time performance.

  + added the ability to have a cluster node's name be different

	from it's uname -n

  + Moved where CTS gets installed to /usr/lib/heartbeat/cts

  + Big improvements to the CTS README from IBM test labs in Austin.

  + bug fixes to the WTI NPS power switch

  + new client API calls:

	return arbitrary configuration parameters

	return current resource status

  + Added a new clplumbing function: mssleep()

  + added new capabilities for supporting pseudo-resources

  + added new messages which come out after initial takeover is done

	 (improves CTS results)

  + LOTS of documentation updates.

  + fixed a security vulnerability

  + fixed a bug where heartbeat would shut down while in the middle

	of processing resource movement requests.

  + changed compilation flags to eliminate similar future security

	issues

  + went to even-more-strict gcc flags

  + fixed several "reload" bugs.  Now reload works ;-)

  + fixed STONITH bug when other node never heard from.

  + Minor bug fixes (cleaned up corrupted message)

  + Two different client API bugs fixed.

  + changed the configure script to test which warning flags are

	supported by the current gcc.

  + enhanced the API test program to test new capabilities...

* Wed May 21 2003  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.1.1

  + Significant restructuring of the processes in heartbeat

  + Added a new feature to hb_standby which allows you to give up

	  any specific category of resources:  local, foreign, or all.

	The old behavior is "all" which is the default.

	This allows you to put a nice_failback cluster into

	  an active/active configuration

  + Enhancements to the ServeRAID code to make it work with the new

    (supported) version of IPSSEND from the ServeRAID folks...

  + Added STONITH code for the Dell remote access controller

  + Fixed a major bug which kept it from taking over correctly after 246

	days or so

  + Fixed a major bug where heartbeat didn't lock itself into memory

	properly

  + Added new ping node timeout directive "deadping"

  + Made sure heartbeat preallocated stack and heap, and printed a

	message if we allocate heap once we're started up...

  + Minor heartbeat API bug fixes

  + Minor documentation fixes

  + Minor fix to allow IP addresses with /32 masks...

  + Fixed a timing window for !nice_failback resource acquisition

  + Added several CCM bug fixes

  + Made the APC smart UPS ups code compatible with more UPSes

  + Fixed a bug in respawn

  + Enhanced internal checking for malloc errors...

  + Added IP alias search optimization from Sean Reifscheneider

* Wed Mar 19 2003  Alan Robertson <alanr@unix.sh> (see doc/AUTHORS file)

+ Version 1.0.2:

  + Fixed comment errors in heartbeat init script to allow it to run on RH 8.0